RE: [WIP]Vertical Clustered Index (columnar store extension) - take2

From: "Aya Iwata (Fujitsu)" <iwata(dot)aya(at)fujitsu(dot)com>
To: 'Timur Magomedov' <t(dot)magomedov(at)postgrespro(dot)ru>
Cc: Peter Smith <smithpb2250(at)gmail(dot)com>, "pgsql-hackers(at)lists(dot)postgresql(dot)org" <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Tomas Vondra <tomas(at)vondra(dot)me>, Japin Li <japinli(at)hotmail(dot)com>, shveta malik <shveta(dot)malik(at)gmail(dot)com>, "Hayato Kuroda (Fujitsu)" <kuroda(dot)hayato(at)fujitsu(dot)com>, 'Alexandre Felipe' <o(dot)alexandre(dot)felipe(at)gmail(dot)com>, Álvaro Herrera <alvherre(at)kurilemu(dot)de>
Subject: RE: [WIP]Vertical Clustered Index (columnar store extension) - take2
Date: 2026-02-17 12:09:46
Message-ID: TYWPR01MB8901CEA325DDC1F6B4B2489FEA6DA@TYWPR01MB8901.jpnprd01.prod.outlook.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi Timur-san

Thank you for your comment.

> 1. Despite performance and other issues that can be fixed anyway, VCI
> has great user API. Index doesn't always mean btree index, it is just a
> database structure for faster data retrieval. In case of VCI it is
> columnar storage for faster data retrieval. It is clear SQL syntax to
> put some columns (not the whole table, we can choose specific columns!)
> into columnar storage. Users are aware of the fact that indexes can
> slow down insertions, no surprises here. No need to change queries,
> just add VCI index for some columns that are scanned heavily without
> putting the whole table into columnar storage.

Yes, VCI allows it to be built only for certain columns
if we choose the IAM-based approach.
I feel that if there are no implementation issues, IAM-based may be a good
choice.

> 2. VCI has own Change Data Capture method. It doesn't use triggers
> either logical replication for CDC, it uses IAM and homegrown hooks in
> heapam.c. Probably VCI-style CDC could be separated to its own patch
> for review and used in other extensions too.

Could you elaborate a bit more on your thoughts regarding this?
For example, what kind of hook functionality is thought to be useful?
It's helpful if we decide hooks are really needed after some research,
and we have to propose them separately.

Best regards,
Aya Iwata

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Aditya Kamath 2026-02-17 12:42:13 RE: AIX support
Previous Message Nazir Bilal Yavuz 2026-02-17 12:04:45 Re: BUG #19095: Test if function exit() is used fail when linked static